In the current era, with the emergence of the so-called digital identity, identity theft has gained new momentum. In view of this scenario, new user authentication mechanisms were developed. However, these processes are generally applied only in the system startup. Consequently, the system would be vulnerable in case the user leaves the workstation and do not end the session.With the aim of mitigating this problem, more recently, a new concept of intrusion detection based on the user's behavior has emerged. In this way, any observed event which deviates from the user's regular behavior is considered a potential intrusion, preventing intruders from using the system. The definition of the user profile may take into consideration a number of aspects, however, in this project, the proposal is to focus on the study of the keystroke dynamics, which involves the analysis of the typing rhythm of the user.Nevertheless, data from keystroke dynamics is noisy and, therefore, the pattern recognition task in this scenario became complex. Among the available tools in Computational Intelligence to solve this problem, immune algorithms deserve to be highlighted due to the success observed in several applications. This project has the goal to propose a method of applying artificial immune systems for pattern recognition in keystroke dynamics. Its performance will be compared to other pattern recognition techniques, considering the state of the art of the areas related to the project.
News published in Agência FAPESP Newsletter about the scholarship: